Currently, in order to enhance the computing power of server, multiple chips and multiple memory banks are generally arranged on a motherboard for the server, and the multiple chips are arranged in sequence and the memory banks are arranged on both sides of the chips respectively. Heat dissipation is usually required to be performed on the chips and the memory banks during operation of the server.
Presently, a main heat dissipation method for the multiple chips and the multiple memory banks in the server is described below: taking two chips as a group, the two chips in the group are arranged along an airflow direction in sequence, the airflow firstly flows through a heat sink of the chip at a front end, and takes away heat of the heat sink, and then the airflow enters the chip at a rear end directly. In this way, after the airflow flows through the chip at the front end, a temperature of the airflow rises to a certain degree, and a cooling intensity is greatly decreased when the airflow enters the chip at the rear end, hence, a service life of the chip at the front end is significantly longer than that of the chip at the rear end. Therefore, the conventional heat dissipation method results in nonuniform heat dissipation between the chips in the server.